Multiscale drivers of carabid beetle (Coleoptera: Carabidae) assemblages in small European woodlands
Abstract
Aim
The spatio‐temporal connectivity of forest patches in lowland agricultural landscapes and their age matter to explain current biodiversity patterns across regional as well as biogeographical extents, to the point that their effect exceeds the one of macroclimate for plant diversity in the understorey of temperate forests. Whether this remains true for other taxonomic groups is still largely unknown. Yet, this relative influence has important consequences for ecosystem functioning and the delivery of ecosystem services. Focusing on carabid beetle assemblages, we assessed the relative importance of macroclimatic, landscape and patch attributes in driving local species richness (α‐diversity) and species dissimilarity between patches (β‐diversity).
Location
Deciduous forest patches in seven regions along a 2,100‐km‐long latitudinal gradient across the European temperate forest biome, from southern France to central Sweden.
Methods
We sampled 221 forest patches in two 5‐km × 5‐km landscape windows with contrasting management intensities. Carabid beetles were classified into four habitat‐preference guilds: forest‐specialist, forest‐generalist, eurytopic and open‐habitat species. We quantified the multi‐level environmental influence using mixed‐effects models and variation partitioning analysis.
Results
We found that both α‐ and β‐diversity were primarily determined by macroclimate, acting as a large‐scale ecological filter on carabid assemblages among regions. Forest‐patch conditions, including biotic and abiotic heterogeneity as well as patch age (but not patch size), increased α‐diversity of forest species. Landscape management intensity weakly influenced α‐diversity of forest species, but increased the number of non‐forest species in forest patches. Beta diversity of non‐forest species increased with patch heterogeneity and decreased with landscape management intensity.
Main conclusions
Our results highlight the leading role of broad macroclimatic gradients over local and landscape factors in determining the composition of local carabid communities, thereby shedding light on macroecological patterns of arthropod assemblages. This study emphasizes the urgent need for preserving ancient forest patches embedded in agricultural landscapes, even the small and weakly connected ones.
1 INTRODUCTION
The cover of present‐day forests in European lowlands mainly consists of forest patches of various size, age, tree species composition, and degree of isolation, that are embedded in more or less intensively managed agricultural matrices (Estreguil et al., 2013). Forest loss and fragmentation are widely acknowledged as major threats to biodiversity and ecosystem functioning through their effects on habitat availability, quality and connectivity (Fahrig, 2003; Foley et al., 2005; Haddad et al., 2015). In particular, increased fragmentation reduces the amount of forest interior habitat that is available for habitat specialist species (Pfeifer et al., 2017). Therefore, patch‐level biotic homogenization (i.e., loss of genetic, taxonomic and functional diversity of species assemblages) due to edge effects is generally observed, which increases in importance as patch size decreases (McKinney & Lockwood, 1999; Olden & Rooney, 2006). At the same time, forest edges can offer suitable conditions for a number of animal and plant generalist species originating from neighbouring, more open or/and disturbed habitats (e.g., Cousins & Eriksson, 2002; De Smedt et al., 2018; González et al., 2017), thereby potentially enhancing the delivery of multiple ecosystem services in small forest patches (Kütt et al., 2018; Valdés et al., 2020). Another consequence of forest loss and fragmentation is that remnant forest patches are increasingly distant from each other. Moreover, the landscape matrix into which forest patches are embedded may be more or less permeable to species movements. Patch isolation and matrix composition both determine landscape connectivity, that is, the degree to which landscape mosaics facilitate movement of organisms and their genes between distinct forest patches (Ricotta et al., 2000; Taylor et al., 1993; Tischendorf & Fahrig, 2000). In particular, a low management intensity of the agricultural matrix (e.g., with grasslands, which are expected to be more permeable than intensively cultivated croplands) and the presence of ecological corridors (e.g., hedgerows, which are thought to facilitate dispersal of forest species among forest patches) will increase colonization rates, while decreasing local extinction rates, in many taxa, including arthropods (e.g., Baudry & Burel, 2019; Seibold et al., 2019). Reduced connectivity between forest patches represents a major threat to species that are forest specialists, including arthropods, and which often exhibit low dispersal abilities (Aviron et al., 2018). As a result, colonization rate of forest patches may be disproportionately low compared to extinction rate and metacommunity functioning could be disrupted. In the case of neutral (Hubbell, 2005) or niche differentiation processes (MacArthur & Levins, 1967; Maire et al., 2012), such a disequilibrium can ultimately lead to an increased dissimilarity in species composition between forest patches, due to species loss or/and species replacement (Hendrickx et al., 2009), thereby increasing the magnitude of difference between the regional species pool and composition of local assemblages. On the contrary, in the case of landscape‐level filtering, species with similar trait values might be selected, leading to species homogenization at the landscape scale (Keddy, 1992; Maire et al., 2012).
In addition to the degree of spatial connectivity, temporal connectivity through the age of forest patches has repeatedly been shown to influence community composition, with ancient forest patches (i.e., patches that have continuously existed for centuries) harbouring a higher number of forest‐specialist species with limited dispersal abilities than recent forest patches (i.e., patches that have established on former agricultural lands in recent decades) (Debnár et al., 2016; Dekoninck et al., 2005; Desender et al., 2005; Lelli et al., 2019). Ancient forest patches usually support more diverse plant assemblages than recent ones, given their higher habitat heterogeneity (Brunet, 2007; Burrascano et al., 2018) and the longer time period available for accumulating forest specialist species (i.e., the species–time relationship; Almoussawi et al., 2020; Valdés et al., 2020). However, compared to plant species, arthropod species have received little attention with respect to species richness in ancient versus recent forest patches (Schowalter, 2017). Besides, to the best of our knowledge, only a few studies have compared the diversity patterns of such arthropod assemblages in different landscape mosaic contexts (e.g., De Smedt et al., 2018; Hendrickx et al., 2009).
Among major arthropod groups living in forest ecosystems, carabid beetles (Coleoptera: Carabidae) are important for ecosystem functioning and biocontrol (Kotze et al., 2011; Kromp, 1999; Thiele, 1977), and are often used as bioindicators of good forest management practices ensuring healthy and sustainable forest systems (e.g., Pearce & Venier, 2006; Rainio & Niemelä, 2003). Forest carabid species are highly sensitive to reduction of forest patch size, as smaller patches exhibit lower habitat quality and greater edge effects than larger patches. Forest fragmentation usually induces a complete replacement of large specialist species characterized by their low mobility by small, generalist, and highly mobile species coming from surrounding habitats (Do & Joo, 2013; Jung et al., 2018; Lövei et al., 2006; Niemelä, 2001; Rainio & Niemelä, 2003). Variations in carabid composition and diversity between mature and young forest stands have previously been shown (e.g., Fountain‐Jones et al., 2015), as well as in function of the forest management (e.g., Barkley et al., 2016; Sroka & Finch, 2006). Surprisingly, the role of historical continuity (i.e., forest patch age) has been largely neglected so far (but see Assmann, 1999; Desender et al., 2005; Hülsmann et al., 2019).
Local climatic conditions (Park et al., 2017), as well as current large‐scale macroclimatic gradients (e.g., Heino & Alahuhta, 2015) and the biogeographical history of the continent since the Pleistocene glaciations (Calatayud et al., 2016, 2019), are widely acknowledged as important drivers of local carabid species assemblages. Together with local and landscape features, macroclimatic factors have been shown to affect carabid assemblages individually, but their relative importance is still unknown, despite their crucial importance for understanding all processes underlying local arthropod assemblages. Whether common patterns are observed across various carabid guilds remains an unanswered question. Given the importance of small deciduous forest patches as biodiversity reservoirs in agricultural landscapes (De Blois et al., 2002) and in delivering multiple ecosystem services to human societies (Decocq et al., 2016), understanding community assembly processes in these habitats embedded in patchy landscapes is particularly relevant to conservation ecology and sustainable landscape planning.
Here we aim at explaining patch‐scale species richness (hereafter α‐diversity) and magnitudes of difference between landscape‐scale species pool and patch‐scale carabid assemblages (hereafter β‐diversity) in small forest patches embedded in contrasting agricultural landscapes along a broad macroclimatic gradient across temperate Europe. We separated carabid assemblages into four different guilds based on habitat preferences, namely: forest specialists, forest generalists, open‐habitat species and eurytopic species. The first two guilds group ‘forest species’, whilst the latter two group ‘non‐forest species’. More specifically, we address the following research hypotheses: (a) macroclimatic conditions act as the main ecological filter on all carabid beetle assemblages; (b) landscape management intensity decreases local species richness (α‐diversity), and increases the magnitude of difference between the landscape species pool and local assemblages (β‐diversity) of forest species (particularly for specialists), whilst the reverse is true for non‐forest species (particularly for open‐habitat species); and (c) patch size and patch age both positively affect α‐diversity, especially for forest species, and forest specialists in particular. To test these hypotheses, we quantified the relative importance of potential environmental drivers acting at different spatial scales (patch, landscape, continent) on local carabid assemblages, using a large, dedicated field survey that encompasses seven regions distributed along a transect running from southern France to central Sweden.
2 METHODS
2.1 Study sites
We collected data across a total of 221 deciduous forest patches, distributed among seven European regions. Regions were distributed along a south‐west to north‐east gradient of c. 2,100 km across the European temperate forest biome (Figure 1). In each region, we selected two 5‐km × 5‐km landscape windows differing by their degree of landscape permeability (see Valdés et al., 2015 for more details). The first window was characterized by forest patches embedded in an intensively cultivated open field landscape (hereafter ‘open field’). The second window contained forest patches better connected by woody corridors (e.g., hedgerows) within a less intensively managed landscape, dominated by grasslands and small crop fields (hereafter ‘bocage’). A detailed and updated list of landscape and macroclimatic variables associated with each landscape window can be found in Vanneste et al. (2019).
For each landscape window, we computed area, perimeter and age of all forest patches using digitized maps (one contemporary map at a scale of 1:25,000 and historical maps from the 17th, 18th, 19th and 20th centuries) in a geographic information system (GIS; ArcGIS 9.3, ESRI, Redlands, CA). Patches were subsequently distributed among four classes according to their area and age: small (<median patch area value for the focal window) and recent (<150 years); large (> median patch area value for the focal window) and recent; small and ancient (> 150 years); and large and ancient.
2.2 Carabid sampling and habitat preference
Whenever possible, we selected four forest patches (i.e., four repetitions) per level of patch size–age combination (n = 4) and per landscape window (n = 14). A perfect balanced design was achieved in five out of seven regions, to finally include a total of 221 forest patches to trap carabid beetles (i.e., 16 forest patches in each window, except 14 in the open field window in eastern Germany and 15 in the open field window in southern Sweden). For this purpose, we used 10‐cm‐diameter pitfall traps installed for 14 consecutive days in both spring (c. April) and summer (c. August) 2013. Traps were filled with 200 mL of a 50% conservative solution of ethylene‐glycol and a few drops of detergent, and protected from litter and rainfall by aluminium roofs. A total of four pairs of traps were placed in each patch as follows. The first pair, consisting of two traps separated by a plastic barrier (100 cm long, 18 cm high), was installed in the inner part of a south‐facing edge (or, when not possible, first a west‐, and then an east‐facing edge was chosen). This setup was replicated 5 m apart along the same edge. A third pair was installed at the barycentre of the forest patch (except in eastern Germany, where all traps were located in the edge), and similarly replicated (fourth pair). The plastic barrier was always parallel to the selected forest edge. To make data comparable among the seven studied regions, and because of the latitudinal climatic gradient covered by our study, the two sampling sessions carried out in each region started when local growing degree hours (GDH; Graae et al., 2012) reached values of c. 10,000 and 20,000 °C h, respectively. Following trap collection, carabid beetles were sorted in the lab in a 70% ethanol solution and identified to the species level following Jeannel (1941, 1942). Species names follow Fauna Europaea (de Jong et al., 2014). Data from all pitfall traps of a given patch and from the two trapping sessions were pooled in all subsequent statistical analyses.
Species were distributed among four guilds, according to their habitat preference and using knowledge from the scientific literature (Bräunicke & Trautner, 2009; Gaublomme et al., 2008; Hůrka, 1996; Sadler et al., 2006): forest‐specialist species, limited to stable, mature forest stands; forest‐generalist species, occurring in any type of forest stand, in ancient as well as recent forest; open‐habitat species, associated with non‐forest habitats such as grasslands and arable lands; and eurytopic species, occurring in open habitats and tolerating transiently forest habitats. We could not assign only one species (Oodes helopioides, n = 1 individual) to any group because of a lack of information in the literature.
2.3 Environmental variables
Three groups of explanatory variables (patch, landscape and macroclimatic variables) were derived from field observations, historical archives or global climatic layers, for and around each forest patch using a GIS.
2.3.1 Macroclimatic variables
To assess the influence of macroclimate on species diversity, we extracted 10 candidate bioclimatic variables from the WorldClim global database (1‐km resolution, http://www.worldclim.org), and averaged each variable for each forest patch using all 1‐km2 pixels intersecting it. Four macroclimatic variables were retained for further analyses, based on a principal component analysis (PCA; see Valdés et al., 2015), namely maximum temperature of the warmest month (MaTWm; BIO5); minimum temperature of the coldest month (MiTCm; BIO6); precipitation of the wettest month (PWm; BIO13); and precipitation of the driest month (PDm; BIO14). The selection was made in such a way as to minimize the correlation between variables and to maximize the correlation with the PCA axes. Our variables were correlated (Pearson’s r) as follows with the PC1 (67.0% of explained variance) and PC2 (17.2%): BIO5 (PC1: r = −.64; PC2: r = −.12), BIO6 (PC1: −.93; PC2: 0.34), BIO13 (PC1: −.52; PC2: −.85) and BIO14 (PC1: −.94; PC2: −.04).
2.3.2 Landscape variables
Landscape variables were computed for five concentric ‘doughnut’‐like buffers of increasing width around each forest patch: 50, 100, 250, 500 and 1,000 m width. We used Corine Land Cover 2006 (Büttner & Kosztra, 2007) to map the distribution of woodland, cropland and grassland. We digitized hedgerows from aerial photographs. As proposed by Martin and Fahrig (2012) and Fahrig (2013), we considered composition‐based measurements of spatial isolation for each forest patch, by calculating the proportion of each cover type and the hedgerow density within each buffer.
2.3.3 Patch‐scale attributes
We included patch area and patch age as covariates in our analyses, to account for both the species–area and species–time relationships (Rosenzweig, 1995). We took the coefficient of variation in elevation values (CVe) within a given forest patch using the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) Global Digital Elevation Map at 30‐m resolution (see Valdés et al., 2015 for further details), as a proxy for heterogeneity of abiotic conditions (including microclimate, soil conditions and light availability) (Graae et al., 2018; Lenoir et al., 2017). Finally, we computed a dissimilarity index in understorey plant species composition within each forest patch (i.e., intra‐patch β‐diversity; see Valdés et al., 2015 for details on computation), separately for forest plant specialists and generalists following distinction criteria as in Valdés et al. (2015).
2.4 Data analysis
All statistical analyses were performed in R 3.5.0 (R Core Team, 2018).
2.4.1 Multiple‐site dissimilarities
To evaluate whether between‐patch differences in carabid composition were due to replacement of species or to nestedness (since the same sampling effort was applied irrespective of patch size), we computed multiple‐site dissimilarities, separating the turnover and nestedness‐resultant components of overall Sørensen‐based multiple‐site dissimilarity as proposed by Baselga (2010, 2012). Dissimilarities were evaluated at two different scales: (a) between‐patch dissimilarity within each landscape window, and (b) between‐region dissimilarity along the European gradient. We used the ‘betapart’ package (Baselga et al., 2018).
2.4.2 Species diversity
We computed patch‐level α‐diversity, defined as the total number of species trapped per forest patch; landscape‐level γ‐diversity, defined as the total number of species retrieved in all patches of a given landscape window; and β‐diversity, defined as the magnitude of difference between the landscape‐level species pool (γ‐diversity) and the composition of the local assemblage (α‐diversity): (γ–α)/γ. Diversity values were calculated separately for each of the four carabid guilds, using the ‘vegan’ package (Oksanen et al., 2018).
2.4.3 Model selection
The effects of patch, landscape and macroclimatic attributes on α‐diversity and β‐diversity were quantified using generalized linear mixed‐effects models (GLMMs) and the ‘lme4’ package (Bates et al., 2015). Negative binomial distribution models (including an observation level factor to account for overdispersion) were used for patch‐level α‐diversity. Gaussian error distributions with an identity link were used for β‐diversity. Distribution families were chosen using the functions descdist and fitdist of the ‘fitdistrplus’ package (Delignette‐Muller & Dutang, 2015). To address possible multicollinearity issues between macroclimatic, landscape and patch attributes, we computed pairwise Pearson correlation tests between all continuous variables, and considered a threshold of .70 to consider two independent predictors as correlated (Dormann et al., 2013; Supporting Information Appendix S1). All explanatory variables were standardized prior to analyses by subtracting the mean and dividing by the standard deviation. We included ‘region’ and ‘window type’ (nested within ‘region’) as random intercept terms in all models to account for the hierarchical structure of the sampling design and potential spatial autocorrelation between assemblages occurring in the same region and landscape window.
In a first step, we selected the most relevant spatial scale (from 50 to 1,000 m) to consider for each landscape variable (i.e., hedgerow density, grassland, crop, and forest relative proportion) at each diversity level (α‐ and β‐diversity) and for each guild. For each landscape variable, we ran a model at each spatial scale (one variable at a time) and retained the one scale from the model with the lowest Akaike information criterion (AIC; Burnham & Anderson, 2004).
In a second step, we constructed complete models regrouping all explanatory variables (see ‘Environmental variables’): first order (BIOX) and second order (BIOX2; to allow for nonlinear macroclimatic variation) macroclimatic variables (n = 8); landscape variables (n = 4); and patch variables (n = 5). Because of the large number of variables, we could not compare candidate models with all possible combinations of variables. We thus performed a backward stepwise selection of fixed explanatory variables: the full model was simplified step‐by‐step by removing the most non‐significant explanatory variables based on likelihood‐ratio tests. The maximum likelihood (ML) estimation was used during model selection, and then restricted maximum likelihood (REML) was used for fitting the final model.
2.4.4 Variation partitioning
We ran a variation partitioning procedure including all response variables to quantify the respective explanatory power of the three groups of explanatory variables (patch, n = 5 variables; landscape, n = 4; and macroclimatic, n = 8) according to Legendre and Legendre (2012) and Valdés et al. (2015). We constructed GLMMs (α‐diversity) or LMMs (β‐diversity) including combinations of one, two or three variable groups plus the random effects, and determined the proportion of variation explained by the fixed variables (marginal R2; 
) according to Nakagawa and Schielzeth (2013). We quantified the proportion of this variation explained by the unique and shared contribution of patch, landscape and macroclimatic variables, and expressed this relative to the total amount of variation explained by the fixed factors in the global model (containing all three groups of explanatory variables).
3 RESULTS
A total of 35,072 individuals corresponding to 159 species were collected from the 221 forest patches. Seven highly frequent species were retrieved in more than 100 patches [Abax parallelepipedus (n = 2,443 individuals), Carabus hortensis (n = 968), Carabus nemoralis (n = 1,206), Nebria brevicollis (n = 1,605), Pterostichus melanarius (n = 6,583), Pterostichus niger (n = 2,095) and Pterostichus oblongopunctatus (n = 2,926)]. Ten species (n = 5,270 individuals) were forest specialist (FS), 24 species (n = 12,350) forest generalist (FG), 51 species (n = 14,803) eurytopic (EU) and 72 species (n = 2,613) open‐habitat (OH) (see Supporting Information Appendix S2 for the full list of species together with their associated habitat‐preference guild and total abundance).
We observed important dissimilarities in species composition among regions along the European gradient, irrespective of the guild considered (Sørensen dissimilarity: .62–.72). These were mainly driven by species turnover (Simpson dissimilarity: .50–.61; Figure 2). Similar results were found among patches within each landscape window and for each guild, except for forest‐specialist species, for which the nestedness component of Sørensen dissimilarity almost equalled the turnover component (Figure 2).
3.1 Species richness (α‐diversity)
Species richness varied along the latitudinal gradient for all guilds, with a trend toward more species at intermediate than at extreme latitudes along our European gradient (Figure 3). In most regions, open‐habitat species richness was (or tended to be) higher in open field than in bocage windows (Figure 3d). We found no significant difference in mean species richness of the three other guilds between open field and bocage windows, irrespective of the region considered (except in northern France: fewer eurytopic species in bocage than open field window; Figure 3c).
Based on the marginal and conditional R2 difference derived from mixed‐effects models, and contrary to other guilds, variance in forest‐specialist species richness was mostly explained by random factors (i.e., region and window type nested in region; Figure 4a–d; Table 1a). Among all groups of explanatory variables, the unique effect of macroclimate explained the largest proportion of local species richness (α‐diversity) for all guilds (89.8, 84.3 and 75.5% of the explained variance for forest‐specialist, eurytopic and open‐habitat species, within the fixed part of the model, respectively; Figure 4a,c,d) except forest generalists. The percentage of variation explained by patch‐scale attributes and landscape variables was negligible. Variation in forest‐generalist species richness was better explained by the unique (50.3%) and total (86.5%) effect of patch‐scale attributes, than by unique (19.0%) and total (49.2%) effects of macroclimatic variables (Figure 4b).
) and conditional R2 (
) of the global models are shown for each response variable. Variation partitioning was based on generalized (GLMM; α‐diversity) or linear mixed‐effect models (LMM; β‐diversity) with the predictor variables as fixed effects and ‘region’ as well as ‘window type’ (nested within ‘region’) as random intercept terms
), conditional R2 () and parameter estimates (± standard error of the regression) for the best model. For landscape attributes, we also indicated the selected spatial scale. The significance of each explanatory variable was obtained from a χ2 test, and is indicated as: ***p < .001; **p < .01; *p < .05; p < .10 (significant estimates are in bold). All variables were standardized prior to analysis. Spaces with a dash indicate that the variable was not included in the best model. Abbreviations are explained under the table
| Explanatory variables | Forest specialist | Forest generalist | Eurytopic | Open habitat | |||||
|---|---|---|---|---|---|---|---|---|---|
| Estimate (± SE) | Estimate (± SE) | Estimate (± SE) | Estimate (± SE) | ||||||
| (a) α‐diversity | |||||||||
| = .361
|
= .653
|
= .582
|
= .438
|
||||||
| = .825
|
= .653
|
= .582
|
= .638
|
||||||
| Macroclimate | MaTWm | – | −0.06 (± 0.05) | −0.17 (± 0.07)* | −0.39 (± 0.20) | ||||
| MaTWm^2 | – | 0.32 (± 0.04 )*** | ‐ | 0.22 (± 0.10)* | |||||
| MiTCm | −1.11 (± 0.36)** | −0.31 (± 0.06)*** | −0.07 (± 0.12) | 1.31 (± 0.33)*** | |||||
| MiTCm^2 | – | −0.55 (± 0.05)*** | −0.13 (± 0.03)*** | −0.62 (± 0.13)*** | |||||
| PWm | – | – | −0.11 (± 0.03)*** | – | |||||
| PWm^2 | – | – | – | – | |||||
| PDm | – | – | −0.21 (± 0.09)* | −1.19 (± 0.20)*** | |||||
| PDm^2 | – | – | −0.18 (± 0.04)*** | 0.31 (± 0.11)** | |||||
| Landscape | Hedgerow length | – | – | – | – | ||||
| Crops (%) | – | – | – | – | |||||
| Grassland (%) | – | – | – | 100 m | −0.15 (± 0.05)** | ||||
| Forest (%) | – | ‐ | 500 m | −0.14 (± 0.03)*** | 500 m | −0.16 (± 0.07)* | |||
| Patch | Age (recent versus old) | – | −0.11 (± 0.05)* | – | – | ||||
| Area | – | – | – | ‐ | |||||
| CVe | 0.12 (± 0.04)** | – | −0.10 (± 0.02)*** | −0.19 (± 0.05)*** | |||||
| β‐div. generalist plants | – | −0.68 (± 0.27)* | – | 0.19 (± 0.06)** | |||||
| β‐div. forest spec. plants | – | 0.95 (± 0.38)* | – | −0.14 (± 0.04)** | |||||
| (b) β‐diversity | |||||||||
| = .284
|
= .153
|
= .285
|
= .096
|
||||||
| = .653
|
= .359
|
= .286
|
= .232
|
||||||
| Macroclimate | MaTWm | – | −0.08 (± 0.02)* | 0.02 (± 0.01)* | – | ||||
| MaTWm^2 | – | – | – | – | |||||
| MiTCm | 0.38 (± 0.12)* | 0.09 (± 0.04)* | – | – | |||||
| MiTCm^2 | – | – | – | – | |||||
| PWm | – | – | 0.03 (± 0.01)** | – | |||||
| PWm^2 | – | – | −0.03 (± 0.01)* | – | |||||
| PDm | −0.18 (± 0.09). | – | 0.04 (± 0.01)** | – | |||||
| PDm^2 | 0.18 (± 0.06)** | – | 0.02 (± 0.01)* | – | |||||
| Landscape | Hedgerow length | – | 50 m | 0.02 (± 0.01)** | – | – | |||
| Crops (%) | – | – | 250 m | −0.04 (± 0.01)*** | 500 m | −0.02 (± 0.01)* | |||
| Grassland (%) | – | – | 1,000 m | −0.04 (± 0.01)** | – | ||||
| Forest (%) | – | 500 m | 0.03 (± 0.01)* | – | 50 m | −0.02 (± 0.01)* | |||
| Patch | Age (recent versus old) | 0.07 (± 0.02)** | 0.03 (± 0.01)* | – | – | ||||
| Area | – | – | – | – | |||||
| CVe | −0.06 (± 0.01)*** | – | 0.03 (± 0.01)*** | 0.03 (± 0.01)*** | |||||
| β‐div. generalist plants | – | 0.03 (± 0.01)** | – | −0.03 (± 0.01)** | |||||
| β‐div. forest spec. plants | – | −0.03 (± 0.01)*** | – | 0.03 (± 0.01)** | |||||
Note
- MaTWm (BIO5) = maximum temperature of the warmest month; MiTCm (BIO6) = minimum temperature of the coldest month; PWm (BIO13) = precipitation of the wettest month; PDm (BIO14): precipitation of the driest month; CVe = coefficient of variation in elevation.
Local carabid species richness was chiefly affected by temperature, across all guilds (Table 1a). As the maximum temperature during the warmest month (MaTWm) increased, α‐diversity of forest‐generalist and open‐habitat species increased whilst the one of eurytopic species decreased. Alpha diversity of all guilds decreased with increased minimum temperature during the coldest month (MiTCm), except for open‐habitat species, for which α‐diversity peaked in regions with intermediate values along the studied gradient of MiTCm and steeply decreased towards mildest temperatures (Table 1a). Extremes in rainfall conditions only affected eurytopic and open‐habitat species richness (Table 1a). Alpha diversity of eurytopic species decreased with increasing amount of precipitation during the wettest (PWm) and driest (PDm) months. Alpha diversity of open‐habitat species was the lowest in regions with intermediate amounts of precipitation during the driest month and increased towards both extremes of the latitudinal gradient.
Landscape characteristics only affected eurytopic and open‐habitat species richness (Table 1a), with both eurytopic and open‐habitat species richness decreasing with increasing proportion of forest in the landscape (500 m). Alpha diversity of open‐habitat species further decreased with increasing amount of grassland in the landscape (100 m).
Guilds were impacted differently by patch conditions, but none was influenced by patch area (Table 1a). Species richness for forest generalists was higher in ancient than recent forest patches, and was positively affected by local β‐diversity of forest‐specialist plant species (but negatively affected by local β‐diversity of generalist plant species). The effect of topographic variability (CVe) on α‐diversity was positive for forest‐specialist, but negative for eurytopic and open‐habitat species. Richness for open‐habitat species further increased with increasing local β‐diversity of generalist plants and with decreasing local β‐diversity of forest‐specialist plants.
3.2 Magnitudes of differences between local assemblages and the landscape pool (β‐diversity)
Overall, we found high values of β‐diversity among fragments, irrespective of guild, landscape type and region considered (Figure 5). We found the lowest magnitude of difference between the landscape pool and local assemblages for forest specialists, especially at intermediate latitudes (Figure 5a).
Our model explained a low proportion of the observed variation in β‐diversity of carabid species assemblages for all guilds and especially for open‐habitat species ( = .096; = 0.232; Figure 4e–h; Table 1b).
For all guilds, β‐diversity was mostly explained by the pure effect of macroclimatic variables (from 31.9 to 71.7% of the explained variance; Figure 3e–h), except for open‐habitat species (pure effect of patch‐sale attributes: 41.2%). For all guilds except forest‐specialist species, landscape variables had a non‐negligible effect on β‐diversity (pure effect from 13.4 to 22.9%; Figure 4f–h).
The magnitude of difference between the landscape pool and local assemblages was primarily explained by extreme temperature (forest‐specialist, generalist and eurytopic species) and extreme rainfall (forest‐specialist, eurytopic and open‐habitat species) conditions (Table 1b). Beta diversity of forest‐generalist species decreased with increasing maximum temperature during the warmest month (MaTWm), whilst the reverse pattern was observed for eurytopic species. Both forest‐specialist and forest‐generalist β‐diversity increased with minimum temperature during the coldest month (MiTCm). Beta diversity of eurytopic species peaked in regions with intermediate amounts of precipitation during the wettest month (PWm), and increased with the amount of precipitation during the driest month (PDm). The latter also had a positive effect on forest‐specialist β‐diversity.
Landscape characteristics affected β‐diversity of all guilds except forest‐specialist species (Table 1a). Beta diversity of forest‐generalist species increased with hedgerow density (50 m) and the proportion of forest (500 m) in the landscape. Beta diversity of eurytopic species decreased with increasing proportion of crops (250 m) and grassland (1,000 m) in the surrounding landscape. Beta diversity of open‐habitat species decreased with increasing amounts of crops (500 m) and forest (50 m) in the surrounding landscape.
Beta diversity of forest specialists and forest generalists was higher in recent than ancient forest patches (Table 1a). Topographic variability (CVe) had a negative effect on forest‐specialist β‐diversity, but a positive effect on eurytopic and open‐habitat β‐diversity. Beta diversity of forest‐generalist species increased with β‐diversity of generalist plant species in the herb layer but decreased with increasing β‐diversity of forest specialist plants. The opposite pattern was found for open‐habitat β‐diversity.
4 DISCUSSION
In this study, we have quantified the respective importance of forest patch characteristics, landscape features, and macroclimate in driving α‐ and β‐diversities of carabid beetle assemblages both within and between small forest patches embedded in contrasting agricultural landscapes. We showed that both diversity components were primarily determined by macroclimate, likely via the size and composition of regional species pools. Local carabid species assemblages differed because of habitat preferences. While the diversity of non‐forest species (open‐habitat and eurytopic) in forest patches was mostly influenced by landscape features, especially proportion of different land‐cover types, the diversity of forest species (specialist and generalist) was on the other hand positively affected by patch‐scale attributes, such as patch age and heterogeneity in patch properties (abiotic and biotic).
4.1 Macroclimate: a prominent ecological filter
We found support for our first research hypothesis: macroclimatic conditions act as the main ecological filter on carabid beetle diversity patterns, as implied by the high relative importance values of macroclimatic factors in explaining local α‐ and β‐diversity patterns. More precisely, extremes of temperature (cf. MaTWm, MiTCm) and, to a lesser degree, extremes of precipitation (cf. PWm, PDm), emerge as crucial drivers of carabid diversity. This is consistent with earlier studies showing that the strong effect of latitude on carabid species richness is mediated by climate, and then more by temperature than precipitation (e.g., Ernst & Buddle, 2015; Schuldt & Assmann, 2009). In addition, and congruently, regional species pools of all species guilds tend to show a hump‐shaped relationship with lower regional species richness at both extremes of the latitudinal gradient (Supporting Information Appendix S3), while a high species turnover is observed among regions along this gradient (Figure 2). Furthermore, distinct carabid species respond differently to macroclimatic factors, depending on habitat preference.
Forest species mostly respond to annual extreme temperatures, with the α‐diversity of specialist and generalist species increasing as MiTCm and MaTWm decreased and increased, respectively. Cold temperatures (cf. MiTCm) are known to exert strong physiological constraints on overwintering poikilothermous species (Lövei & Sunderland, 1996; Thiele, 1977). The majority of temperate carabid species require winter diapause to complete larval development or gonad maturation and cannot survive environments with too mild winter temperatures, since diapause is not triggered (Kotze et al., 2011). On the other hand, cold induces physiological costs and only a few species have adapted mechanisms to survive very harsh winter conditions (Bale, 1996). High temperature (cf. MaTWm) usually increases activity and development rates (e.g., Forrest, 2016; Taylor, 1963) and throughout the active period of the year, insects that face warm conditions during their ontogenetic development can acclimate to even higher temperatures (Sheikh et al., 2017). Warmer weather conditions may thus promote species movements within landscapes, thereby affecting β‐diversity patterns (here defined as the magnitude of difference between the landscape‐level species pool and the composition of local assemblages).
Among non‐forest species, eurytopic species respond to both temperature (cf. negative effect of both MaTWm and MiTCm on α‐diversity) and precipitation (cf. negative effect of both PWm and PDm on α‐diversity), while open‐habitat α‐diversity rather show a unimodal relationship with MiTCm and PDm. Eurytopic species can live and disperse in most habitats, but may not have developed adaptations for extreme climatic conditions. Moreover, as non‐forest species, eurytopic and open‐habitat species cannot benefit from canopy‐induced microclimatic conditions of the understorey, which is typically moister over the year, cooler in summer, and milder in winter (De Frenne et al., 2019). When the temperature is too high or the environment too dry, these species are thus more prone to enter into summer diapause or quiescence to face unsuitable conditions (Masaki, 1980). Because of reduced activity, these species likely disperse less efficiently, thereby increasing compositional dissimilarity among patches.
As a consequence of current climate change and because of physiological constraints of species, the elevational range of some carabid species has changed in the past 30 years, to track the shifting isotherms (e.g., Moret et al., 2016; Pizzolotto et al., 2014). Movement of species northward (in the Northern Hemisphere) can also be expected, but changes in latitudinal distributions can be hard to predict as the dominant diapause‐inducing cue (photoperiod) will not be affected by climate change while other cues (temperature, moisture) will (Huffeldt, 2020). Species may indeed suffer from mismatched information between the critical photoperiod for diapause induction and other environmental conditions acting on the physiology, development, or behaviour of specific life stages, which could affect their survival (Bale & Hayward, 2010; Tougeron et al., 2019). In addition, by causing species‐specific shifts in phenology, climate change can alter interspecific interactions as well (Damien & Tougeron, 2019).
The legacy of the biogeographical history of Europe (particularly glaciations of the Pleistocene and the subsequent post‐glacial recolonization of northern regions) may blur the pure effect of current climatic conditions in explaining the macroecological diversity patterns we observed. This is particularly true for dispersal‐limited organisms such as carabid beetles (Calatayud et al., 2016, 2019), and may explain the lower species richness recorded in the northern regions.
4.2 Landscape features explain the contribution of non‐forest species to local assemblages
Our results reveal that the composition of the landscape matrix into which forest patches are embedded significantly explains the α‐ and β‐diversity of non‐forest carabid species (cf. eurytopic and open‐habitat guilds). In particular, focal patches surrounded by forest and grassland have a negative effect on species richness for these species. Consistently, the number of open‐habitat species was higher in open field landscapes compared to bocage landscapes. This may be explained by the diversity of crops cultivated in the surrounding fields (e.g., cereals, sugar beet, potato, rapeseed), each associated with a particular pool of species (Holland et al., 2005; Marrec et al., 2015), whilst grasslands may harbour less specific carabid assemblages and benefit from spillover from adjacent habitats, including forests (Duflot et al., 2018; Schneider et al., 2016). Moreover, arable lands may allow beetles to move more easily on the ground than grasslands, particularly because of differences in vegetation density (Thomas et al., 2006). Our results are consistent with previous studies that showed that landscape attributes affect biodiversity in important ways (see Fletcher et al., 2016 for a review). Open‐habitat and eurytopic species use forest habitats as alternative, temporary habitat, but are less competitive there than in more open habitats (e.g., Niemelä et al., 1993). In particular, they enter forest edges to find shelter, to search for food or to seek overwintering sites (Knapp et al., 2019). The decreased magnitude of difference between the landscape‐level species pool and the composition of local assemblages with increasing crop or/and grassland cover in the landscape suggests that these habitats allow a large suite of non‐forest species to enter forest patches. However, the high turnover observed among patches of a given landscape window suggests that they are not present at the same time in most patches.
With the exception of β‐diversity of forest‐generalist species, which increases with hedgerow density and forest cover in the surrounding landscape, landscape factors do not affect the α‐ and β‐diversity patterns of forest‐specialist and generalist carabid species. This result indicates that hedgerows are not always positive for species movement (e.g., Baudry & Burel, 2019). Alternatively, hedgerows in the studied landscapes may be too recent or scarce to serve as high‐quality corridors (Closset‐Kopp et al., 2016; Lenoir et al., 2020) for forest carabid species, which are well known for their very low dispersal capacities (Rainio & Niemelä, 2003). However, hedgerows may allow some forest species to disperse between forest patches as previously suggested (e.g., Baudry & Burel, 2019), increasing the probability that more species will reach forest patches. However, the high nestedness we found for forest‐specialist species compared to other guilds suggests that only a few species are actually able to colonize new forest patches. Increased magnitude of difference between the landscape‐level species pool and the composition of local assemblages in bocage landscapes could then be linked to random local species replacement due to competition, or neutral processes. In summary, we found little support for our second research hypothesis: landscape management intensity only weakly influences the diversity of forest carabid species, but instead increases the number of non‐forest species in small forest patches.
4.3 Forest patch characteristics drive forest carabid species diversity
We found that patch‐scale attributes mostly influence the α‐ and β‐diversity patterns of forest carabid species. Patch characteristics were even the primary drivers of α‐diversity for forest‐generalist species. This includes patch age and various proxies for local environmental heterogeneity. The higher number of forest generalists in ancient compared to recent forest patches likely reflects the accumulation of weaker dispersers over time, according to the species–time relationship (Rosenzweig, 1995), which has already been shown for vascular plant species in the same study system (Valdés et al., 2015). Similarly, forest carabid species are large species, with poor dispersal abilities, and long life cycles (Rainio & Niemelä, 2003; Thiele, 1977). Ancient forests are thought to be more stable, allowing these species to persist (Driscoll & Weir, 2005; Schowalter, 2017). Moreover, ancient forest patches are also thought to exhibit more heterogeneous environmental conditions than recent patches (Honnay et al., 1999; Lawesson et al., 1998; Schowalter, 2017), thereby providing carabid beetles with more potential niches. This is confirmed by the positive effect of plant diversity (cf. intra‐patch β‐diversity of forest specialists) on the species richness of forest generalist carabid beetles, since plant species richness has also been shown to increase with patch heterogeneity (Jamoneau et al., 2011). Carabid beetles have been shown to be very sensitive to variations in forest characteristics, so that their assemblages change during the forestry cycle, according to variation in forest structure and composition (e.g., Butterfield, 1997; Koivula et al., 2002; Magura et al., 2003).
The β‐diversity of forest generalists was influenced by the same patch‐scale factors as α‐diversity, but in the opposite direction. This indicates that the few forest‐generalist species present in a given landscape window tend to occupy all ancient, heterogeneous forest patches of this window. Interestingly, the β‐diversity of forest specialists also decreased as patch age increased, consistent with species accumulation over time. However, local species richness was mostly determined by region and landscape‐window type (i.e., random factors in our models). This result suggests that the colonization of forest patches within landscape windows occurs at random in such fragmented systems where metapopulation dynamics for these dispersal‐limited species are likely disrupted. The positive effect of local topography (another proxy for intra‐patch heterogeneity) on local species richness can be explained by the fact that small‐scale variation in topography offers more thermal variability than flat areas (Lenoir et al., 2013), as well as more micro‐habitats (e.g., various depth and quality of litter, coarse woody debris, snags, different soil texture; Dwyer & Merriam, 1981). These features have already been shown to affect carabid assemblages (Latty et al., 2006; Tyler, 2008), and other arthropods such as the tick Ixodes ricinus (Ehrmann et al., 2017).
It is noteworthy that patch area was not significant in our models, which is likely a consequence of our sampling design, since we applied the same sampling effort irrespective of patch size, using traps that can catch beetles over a very limited area relative to patch size. Nonetheless, recent reviews suggested that small and isolated habitat patches can have similar ecological values as large patches, especially when they encompass the same overall area at the landscape to regional scale (Fahrig, 2017; Wintle et al., 2019).
To a lesser extent, patch‐scale attributes also impact the diversity of open‐habitat species. In particular, the magnitude of difference between the landscape‐level species pool and the composition of their local assemblages was positively affected by local topography and β‐diversity of forest plant specialists. This suggests that open‐habitat species, and, to a lesser degree, eurytopic species (see positive relationship between β‐diversity of eurytopic species and CVe) hardly colonize heterogeneous forests established on complex terrains. Consistently, the number of open‐habitat species (α‐diversity) decreased with increasing vegetation heterogeneity (cf. β‐diversity of forest specialist plants) and increasing elevational heterogeneity (cf. CVe).
Based on these results, we can partly confirm our third research hypothesis: patch age, not patch size (at least when a similar sampling effort is applied irrespective of patch size), increases α‐diversity of forest species, but local abiotic and biotic heterogeneities have an even more positive effect.
5 CONCLUSION
We show that composition of carabid assemblages in small forest patches is strongly driven by macroclimatic conditions (and possibly by confounding historical factors) for all habitat‐preference guilds. This result indicates the importance of considering macroecological factors as the main drivers of local carabid species assemblages. This pattern was yet unknown for most arthropods including carabid beetles, and much needed to improve our knowledge in arthropod macroecological patterns. Indeed, in the current climate change context, integrating climatic information into large‐scale (both temporal and spatial) ecology studies is required. If species facing climate change are not sufficiently plastic or adaptable in terms of their climatic tolerance range, they will have to move northward (in the Northern Hemisphere) or upward in mountainous systems to track the shifting isotherms and survive, and consequently impact local species assemblages.
Species assembly was also importantly altered by patch environmental heterogeneity and age, as previously observed in the same forest networks for other arthropods and plants (De Smedt et al., 2018; Valdés et al., 2015). Yet, we did not find any evidence on an influence of forest size. These results highlight the major need to protect old forest patches embedded in agricultural landscapes, even the small ones. Finally, landscape management intensity weakly influences the diversity of forest carabid species, but increases the number of non‐forest species in small forest patches, while decreasing the turnover of these among patches. These results invalidate our previous assumptions and instead made us consider that the observed patterns, at multiple spatial scales, are likely linked to additional functional traits, rather than species habitat preference only. Congruently, Le Provost et al. (2020) recently showed that mobility, resource‐acquisition, and body‐size traits drive local species assembly through environmental filtering acting at the landscape and patch levels, especially in landscapes suffering high short‐ and long‐term turnover such as agricultural landscapes. Such evidence has not been provided yet on assemblages living in small forest patches embedded in these agricultural landscapes.
ACKNOWLEDGMENTS
We thank Catherine Bataillon, Kent Hansson, Laurent Raison, Sabinea Sigfridsson and Ilka Strubelt for their help during fieldwork, and Willem Proesmans, Laurent Raison and Rieneke Vanhulle for their help for species sorting and identification. This research was funded by the Horizon 2020 ERA‐NET BiodivERsA project smallFOREST, with the national funders Agence Nationale de la Recherche (ANR; France), Ministerio de Economía y Empresa (MINECO; Spain), Forskningsråd för hållbar utveckling (FORMAS; Sweden), Eesti TeadusAgentuur (ETAg; Estonia), Deutsche Forschungsgemeinschaft/Deutsches Zentrum für Luft‐ und Raumfahrt (DFG/ DLR; Germany) and Federaal Wetenschapsbeleid (BELSPO; Belgium) as part of the 2011 BiodivERsA call for research proposals. This paper was written while LM held a PhD fellowship from the « Conseil Régional de Picardie ». PDS holds a post‐doctoral fellowship of the Research Foundation‐Flanders (FWO).
AUTHOR CONTRIBUTIONS
GD, KV, MDi, VLR, LM, BG and PDS designed the study. RM, LM, GD and VLR wrote the manuscript. RM, LM and JoL ran the analyses. LM, PDS, VLR and BG sorted arthropods from the traps and identified the species. LM, VLR, AV, JoL, GD, BG, MDi, PDS, KV, JB, SC, JaL, MDe, MW, SE and JeL collected data in the field and contributed to writing the manuscript. EGM managed the databases and conducted GIS analyses.
REFERENCES
BIOSKETCH
Ronan Marrec is an Assistant Professor at the UMR 7058 CNRS EDYSAN (Ecology and Dynamics of Anthropogenic Systems), Jules Verne University of Picardy, France. As a landscape ecologist he is interested in evaluating the role of the spatial structure and temporal dynamics of landscapes and human activities on biodiversity and ecosystem services. For more information: https://ronanmarrec.weebly.com.
All authors are members of the research project smallFOREST (http://www.u‐picardie.fr/smallforest/uk/), which is devoted to studying biodiversity and ecosystem services of small forest patches in European landscapes.
